Prosody & Sentry
Now that we have access to a Sentry, @MattJ could you estimate how much time you would need to journalise error there from both the production & the pre-production Prosody?
Old accounting Protocol
Link issues together to show that they're related. Learn more.
Activity
- Developer
- Add net.http sender for raven-lua (3h)
- mod_log_sentry to forward errors (4h)
- Deploy to dev server (1h)
- Deploy to production server (1h)
- Matthew Wild changed time estimate to 1d 1h
changed time estimate to 1d 1h
- Jean-Baptiste Pasquier mentioned in issue djangoldp-packages/djangoldp-notification#27 (closed)
mentioned in issue djangoldp-packages/djangoldp-notification#27 (closed)
- Author Owner
@Cyrilthiriet Do you want to fund this?
It'll improve our capacity to spot errors when an user/circle/project save fail between DjangoLDP & Prosody.
Actually we're monitoring the app side, the djangoldp's side but we're missing the Prosody's side to get a complete track of what's happening when someone tell you "I can mention X" or "I didn't received a notification from Y".
125 tokens to fund.
- Developer
I see with @alexbourlier I come back to you.
- Maintainer
@jbpasquier @MattJ I'd tend to follow your recommendation here. If you tell us it is comfortable to have it but we can live without it, then let's not do this right now.
If you tell us that it is a massive pain to debug chat issues without it, then I'm for funding it and having a happy team
- Author Owner
I see that we spend a lot of time/money on chat investigation. Lot of them would have been pretty easier/faster if we used to have the Sentry on Prosody's side too.
While it's certainly not indispensable, I have a strong feeling that while our user base will grow, we'll have more and more debug needed before reaching a stable one and those 9h of Matthew will save us at least the double in less than a month.
- Maintainer
He he, fair enough then
- Developer
If you tell us that it is a massive pain to debug chat issues without it, then I'm for funding it and having a happy team
I think it's a massive pain to debug chat issues right now. Well, ok, not a massive pain. It's not complicated (for someone familiar with Prosody), but it's time-consuming doing manual investigations and it happens too often.
My goals would be:
- Make debugging of issues more accessible - i.e. it should be easier for people other than me to figure out what is going wrong (and when)
- Proactive detection of things going wrong - i.e. not waiting for yet another
"I can't mention @foo"
issue report before knowing there is a problem. Right now users are acting as our Sentry on the chat side...
We've made a bunch of advances (finally!) in improving the link between Prosody and Django. I think with a little more effort we'll be in a fairly good place. The only thing more that I would wish for is a test suite.
For handling production issues though, I would definitely recommend moving ahead with either this issue (i.e. Sentry) or this issue: #206 (moved) (debug tooling).
Both would be nice of course, but if I had to choose one then I think Sentry is more important because it catches a broader range of problems and the "proactive" part. The debug tooling would make everyone's life easier in investigating certain specific and annoying issues that are repeatedly occurring, but that can still be done manually. Hopefully with Sentry in place it will be easier to catch and fix the root causes of such issues.
- Developer
After our discussion @jbpasquier, do we have please a general visual mapping (server and apps ) or similar of what is what and where we are impacted ? Would be very nice to understand. Thx for your help = )
- Author Owner
@Cyrilthiriet Basically, this is how our chat work:
Sentry is our magical tool which allow us to thing things like that (and a lot of other pretty cool things too):
Sentry is triggered whenever something wrong happen, from an user action, a script, anything else.
Sentry is actually configured on the Hubl side and on DjangoLDP side. We're missing the Sentry part on Prosody.
- Maintainer
@MattJ Thanks for your message, it makes it very clear. I'm for funding it then, we'll assess how today or tomorrow.
Same thing regarding the test suite. If you can provide a time estimation, that helps to assess if we can afford it now or not yet
- Author Owner
Debugging tool is 111 tokens
- Please register or sign in to reply
- Jean-Baptiste Pasquier mentioned in issue applications/hubl#873 (closed)
mentioned in issue applications/hubl#873 (closed)
- Developer
Hello @jbpasquier It's ok for FNK to fund the last 125 tokens. We can Go ;) #tokenfnk
Edited by Cyril Thiriet - Author Owner
@MattJ Green light
- Developer
Working on this, made a lot of progress today and hopefully will be wrapped up with code tomorrow.
- Maintainer
@MattJ @jbpasquier Where are we here?
- Developer
@alexbourlier Mea culpa. I dropped the ball on this, will see if I can roll it out to production today.
- Cyril Thiriet changed milestone to %Xmpp Cyril
changed milestone to %Xmpp Cyril
- Jean-Baptiste Pasquier mentioned in issue #240 (closed)
mentioned in issue #240 (closed)
- Developer
The production server is now logging to Sentry with the environment "xmpp-prod"
- Maintainer
Hallelujah my brothers.
Thank you @MattJ
Not to blame anyone but just so he is aware of it: it is technically @Cyrilthiriet's job to poke you on the shoulder as soon as a ball is dropped. I think he is aware of it and taking his marks. Beware of the incoming poking!
Edited by Alex
- Matthew Wild closed
closed
- Cyril Thiriet changed the description
changed the description